diff --git a/internal/services/machinelearning/machine_learning_compute_cluster_resource.go b/internal/services/machinelearning/machine_learning_compute_cluster_resource.go
index aaac3828f166..1fd86e283395 100644
--- a/internal/services/machinelearning/machine_learning_compute_cluster_resource.go
+++ b/internal/services/machinelearning/machine_learning_compute_cluster_resource.go
@@ -155,8 +155,10 @@ func resourceComputeCluster() *pluginsdk.Resource {
 			},
 
 			"subnet_resource_id": {
-				Type:     pluginsdk.TypeString,
+				Type: pluginsdk.TypeString,
+				// NOTE: O+C as you don't have to specify it for Azure to assign one to the cluster
 				Optional: true,
+				Computed: true,
 				ForceNew: true,
 			},
 
@@ -211,28 +213,6 @@ func resourceComputeClusterCreate(d *pluginsdk.ResourceData, meta interface{}) e
 	if !response.WasNotFound(existing.HttpResponse) {
 		return tf.ImportAsExistsError("azurerm_machine_learning_compute_cluster", id.ID())
 	}
-	nodePublicIPEnabled, ok := d.Get("node_public_ip_enabled").(bool)
-	if !ok {
-		return fmt.Errorf("unable to assert type for `node_public_ip_enabled`")
-	}
-
-	subnetResourceID, ok := d.Get("subnet_resource_id").(string)
-	if !ok {
-		return fmt.Errorf("unable to assert type for `subnet_resource_id`")
-	}
-
-	workspaceInManagedVnet := false
-
-	if workspaceModel.Properties != nil &&
-		workspaceModel.Properties.ManagedNetwork != nil &&
-		workspaceModel.Properties.ManagedNetwork.Status != nil &&
-		workspaceModel.Properties.ManagedNetwork.Status.Status != nil {
-		workspaceInManagedVnet = *workspaceModel.Properties.ManagedNetwork.Status.Status == workspaces.ManagedNetworkStatusActive
-	}
-
-	if !nodePublicIPEnabled && subnetResourceID == "" && !workspaceInManagedVnet {
-		return fmt.Errorf("`subnet_resource_id` must be set if `node_public_ip_enabled` is set to `false` or the workspace is not in a managed network")
-	}
 
 	vmPriority := machinelearningcomputes.VMPriority(d.Get("vm_priority").(string))
 	computeClusterAmlComputeProperties := machinelearningcomputes.AmlComputeProperties{
diff --git a/internal/services/machinelearning/machine_learning_compute_cluster_resource_test.go b/internal/services/machinelearning/machine_learning_compute_cluster_resource_test.go
index ad46d54b0757..08a19a83b3ee 100644
--- a/internal/services/machinelearning/machine_learning_compute_cluster_resource_test.go
+++ b/internal/services/machinelearning/machine_learning_compute_cluster_resource_test.go
@@ -65,6 +65,21 @@ func TestAccComputeCluster_complete(t *testing.T) {
 	})
 }
 
+func TestAccComputeCluster_subnetResourceId(t *testing.T) {
+	data := acceptance.BuildTestData(t, "azurerm_machine_learning_compute_cluster", "test")
+	r := ComputeClusterResource{}
+
+	data.ResourceTest(t, r, []acceptance.TestStep{
+		{
+			Config: r.noSubnetResourceId(data),
+			Check: acceptance.ComposeTestCheckFunc(
+				check.That(data.ResourceName).ExistsInAzure(r),
+			),
+		},
+		data.ImportStep(),
+	})
+}
+
 func TestAccComputeCluster_recreateVmSize(t *testing.T) {
 	data := acceptance.BuildTestData(t, "azurerm_machine_learning_compute_cluster", "test")
 	r := ComputeClusterResource{}
@@ -235,6 +250,48 @@ resource "azurerm_machine_learning_compute_cluster" "test" {
 `, template, data.RandomIntOfLength(8))
 }
 
+func (r ComputeClusterResource) noSubnetResourceId(data acceptance.TestData) string {
+	template := r.template_complete(data)
+	return fmt.Sprintf(`
+%s
+variable "ssh_key" {
+  default = "ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQCqaZoyiz1qbdOQ8xEf6uEu1cCwYowo5FHtsBhqLoDnnp7KUTEBN+L2NxRIfQ781rxV6Iq5jSav6b2Q8z5KiseOlvKA/RF2wqU0UPYqQviQhLmW6THTpmrv/YkUCuzxDpsH7DUDhZcwySLKVVe0Qm3+5N2Ta6UYH3lsDf9R9wTP2K/+vAnflKebuypNlmocIvakFWoZda18FOmsOoIVXQ8HWFNCuw9ZCunMSN62QGamCe3dL5cXlkgHYv7ekJE15IA9aOJcM7e90oeTqo+7HTcWfdu0qQqPWY5ujyMw/llas8tsXY85LFqRnr3gJ02bAscjc477+X+j/gkpFoN1QEmt terraform@demo.tld"
+}
+
+resource "azurerm_machine_learning_compute_cluster" "test" {
+  name                          = "CC-%d"
+  location                      = azurerm_resource_group.test.location
+  vm_priority                   = "LowPriority"
+  vm_size                       = "STANDARD_DS2_V2"
+  machine_learning_workspace_id = azurerm_machine_learning_workspace.test.id
+  node_public_ip_enabled        = false
+  description                   = "Machine Learning"
+  tags = {
+    environment = "test"
+  }
+  scale_settings {
+    min_node_count                       = 0
+    max_node_count                       = 1
+    scale_down_nodes_after_idle_duration = "PT30S" # 30 seconds
+  }
+
+  identity {
+    type = "SystemAssigned"
+  }
+
+  ssh_public_access_enabled = false
+  ssh {
+    admin_username = "adminuser"
+    key_value      = var.ssh_key
+  }
+  depends_on = [
+    azurerm_subnet_network_security_group_association.test,
+    azurerm_private_endpoint.test,
+  ]
+}
+`, template, data.RandomIntOfLength(8))
+}
+
 func (r ComputeClusterResource) recreateVmSize(data acceptance.TestData) string {
 	template := r.template_basic(data)
 	return fmt.Sprintf(`
diff --git a/website/docs/r/machine_learning_compute_cluster.html.markdown b/website/docs/r/machine_learning_compute_cluster.html.markdown
index 90f6e2eb2cdd..5753ff83d6f3 100644
--- a/website/docs/r/machine_learning_compute_cluster.html.markdown
+++ b/website/docs/r/machine_learning_compute_cluster.html.markdown
@@ -123,7 +123,7 @@ The following arguments are supported:
 
 * `local_auth_enabled` - (Optional) Whether local authentication methods is enabled. Defaults to `true`. Changing this forces a new Machine Learning Compute Cluster to be created.
 
-* `node_public_ip_enabled` - (Optional) Whether the compute cluster will have a public ip. To set this to false a `subnet_resource_id` needs to be set. Defaults to `true`. Changing this forces a new Machine Learning Compute Cluster to be created.
+* `node_public_ip_enabled` - (Optional) Whether the compute cluster will have a public ip. Defaults to `true`. Changing this forces a new Machine Learning Compute Cluster to be created.
 
 * `ssh_public_access_enabled` - (Optional) A boolean value indicating whether enable the public SSH port. Defaults to `false`. Changing this forces a new Machine Learning Compute Cluster to be created.